SYMBOL INDEX (118 symbols across 18 files)

FILE: dataset/lm_dataset.py
  function pre_processing_chat (line 8) | def pre_processing_chat(conversations, add_system_ratio=0.2):
  function post_processing_chat (line 26) | def post_processing_chat(prompt_content, empty_think_ratio=0.05):
  class PretrainDataset (line 31) | class PretrainDataset(Dataset):
    method __init__ (line 32) | def __init__(self, data_path, tokenizer, max_length=512):
    method __len__ (line 38) | def __len__(self):
    method __getitem__ (line 41) | def __getitem__(self, index):
  class SFTDataset (line 52) | class SFTDataset(Dataset):
    method __init__ (line 53) | def __init__(self, jsonl_path, tokenizer, max_length=1024):
    method __len__ (line 61) | def __len__(self):
    method create_chat_prompt (line 64) | def create_chat_prompt(self, conversations):
    method generate_labels (line 74) | def generate_labels(self, input_ids):
    method __getitem__ (line 92) | def __getitem__(self, index):
  class DPODataset (line 108) | class DPODataset(Dataset):
    method __init__ (line 109) | def __init__(self, file_path, tokenizer, max_length=4096):
    method __len__ (line 118) | def __len__(self):
    method __getitem__ (line 121) | def __getitem__(self, index):
    method generate_loss_mask (line 162) | def generate_loss_mask(self, input_ids):
  class RLAIFDataset (line 181) | class RLAIFDataset(Dataset):
    method __init__ (line 182) | def __init__(self, jsonl_path, tokenizer, max_length=1024):
    method __len__ (line 190) | def __len__(self):
    method create_chat_prompt (line 193) | def create_chat_prompt(self, conversations):
    method __getitem__ (line 208) | def __getitem__(self, index):

FILE: eval_llm.py
  function init_model (line 12) | def init_model(args):
  function main (line 32) | def main():

FILE: model/model_lora.py
  class LoRA (line 6) | class LoRA(nn.Module):
    method __init__ (line 7) | def __init__(self, in_features, out_features, rank):
    method forward (line 17) | def forward(self, x):
  function apply_lora (line 21) | def apply_lora(model, rank=8):
  function load_lora (line 35) | def load_lora(model, path):
  function save_lora (line 45) | def save_lora(model, path):

FILE: model/model_minimind.py
  class MiniMindConfig (line 8) | class MiniMindConfig(PretrainedConfig):
    method __init__ (line 11) | def __init__(
  class RMSNorm (line 96) | class RMSNorm(torch.nn.Module):
    method __init__ (line 97) | def __init__(self, dim: int, eps: float = 1e-5):
    method _norm (line 102) | def _norm(self, x):
    method forward (line 105) | def forward(self, x):
  function precompute_freqs_cis (line 109) | def precompute_freqs_cis(dim: int, end: int = int(32 * 1024), rope_base:...
  function apply_rotary_pos_emb (line 131) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function repeat_kv (line 140) | def repeat_kv(x: torch.Tensor, n_rep: int) -> torch.Tensor:
  class Attention (line 150) | class Attention(nn.Module):
    method __init__ (line 151) | def __init__(self, args: MiniMindConfig):
    method forward (line 169) | def forward(self,
  class FeedForward (line 216) | class FeedForward(nn.Module):
    method __init__ (line 217) | def __init__(self, config: MiniMindConfig):
    method forward (line 228) | def forward(self, x):
  class MoEGate (line 232) | class MoEGate(nn.Module):
    method __init__ (line 233) | def __init__(self, config: MiniMindConfig):
    method reset_parameters (line 248) | def reset_parameters(self) -> None:
    method forward (line 251) | def forward(self, hidden_states):
  class MOEFeedForward (line 288) | class MOEFeedForward(nn.Module):
    method __init__ (line 289) | def __init__(self, config: MiniMindConfig):
    method forward (line 303) | def forward(self, x):
    method moe_infer (line 329) | def moe_infer(self, x, flat_expert_indices, flat_expert_weights):
  class MiniMindBlock (line 352) | class MiniMindBlock(nn.Module):
    method __init__ (line 353) | def __init__(self, layer_id: int, config: MiniMindConfig):
    method forward (line 365) | def forward(self, hidden_states, position_embeddings, past_key_value=N...
  class MiniMindModel (line 376) | class MiniMindModel(nn.Module):
    method __init__ (line 377) | def __init__(self, config: MiniMindConfig):
    method forward (line 392) | def forward(self,
  class MiniMindForCausalLM (line 427) | class MiniMindForCausalLM(PreTrainedModel, GenerationMixin):
    method __init__ (line 430) | def __init__(self, config: MiniMindConfig = None):
    method forward (line 437) | def forward(self,

FILE: scripts/convert_model.py
  function convert_torch2transformers_minimind (line 16) | def convert_torch2transformers_minimind(torch_path, transformers_path, d...
  function convert_torch2transformers_llama (line 36) | def convert_torch2transformers_llama(torch_path, transformers_path, dtyp...
  function convert_transformers2torch (line 65) | def convert_transformers2torch(transformers_path, torch_path):

FILE: scripts/serve_openai_api.py
  function init_model (line 27) | def init_model(args):
  class ChatRequest (line 49) | class ChatRequest(BaseModel):
  class CustomStreamer (line 59) | class CustomStreamer(TextStreamer):
    method __init__ (line 60) | def __init__(self, tokenizer, queue):
    method on_finalized_text (line 65) | def on_finalized_text(self, text: str, stream_end: bool = False):
  function generate_stream_response (line 71) | def generate_stream_response(messages, temperature, top_p, max_tokens):
  function chat_completions (line 114) | async def chat_completions(request: ChatRequest):

FILE: scripts/web_demo.py
  function process_assistant_content (line 71) | def process_assistant_content(content):
  function load_model_tokenizer (line 99) | def load_model_tokenizer(model_path):
  function clear_chat_messages (line 112) | def clear_chat_messages():
  function init_chat_messages (line 117) | def init_chat_messages():
  function regenerate_answer (line 140) | def regenerate_answer(index):
  function delete_conversation (line 146) | def delete_conversation(index):
  function setup_seed (line 197) | def setup_seed(seed):
  function main (line 207) | def main():

FILE: trainer/train_distillation.py
  function distillation_loss (line 24) | def distillation_loss(student_logits, teacher_logits, temperature=1.0, r...
  function train_epoch (line 38) | def train_epoch(epoch, loader, iters, teacher_model, lm_config_student, ...

FILE: trainer/train_dpo.py
  function logits_to_log_probs (line 24) | def logits_to_log_probs(logits, labels):
  function dpo_loss (line 33) | def dpo_loss(ref_log_probs, policy_log_probs, mask, beta):
  function train_epoch (line 54) | def train_epoch(epoch, loader, iters, ref_model, lm_config, start_step=0...

FILE: trainer/train_full_sft.py
  function train_epoch (line 23) | def train_epoch(epoch, loader, iters, start_step=0, wandb=None):

FILE: trainer/train_grpo.py
  function calculate_rewards (line 27) | def calculate_rewards(prompts, responses, reward_model, reward_tokenizer):
  function grpo_train_epoch (line 95) | def grpo_train_epoch(epoch, loader, iters, ref_model, reward_model, rewa...

FILE: trainer/train_lora.py
  function train_epoch (line 24) | def train_epoch(epoch, loader, iters, lora_params, start_step=0, wandb=N...

FILE: trainer/train_ppo.py
  class CriticModel (line 29) | class CriticModel(MiniMindForCausalLM):
    method __init__ (line 30) | def __init__(self, params):
    method forward (line 35) | def forward(self, input_ids=None, attention_mask=None, **kwargs):
  function calculate_rewards (line 44) | def calculate_rewards(prompts, responses, reward_model, reward_tokenizer):
  function ppo_train_epoch (line 119) | def ppo_train_epoch(epoch, loader, iters, old_actor_model, ref_model, ac...

FILE: trainer/train_pretrain.py
  function train_epoch (line 23) | def train_epoch(epoch, loader, iters, start_step=0, wandb=None):

FILE: trainer/train_reason.py
  function train_epoch (line 23) | def train_epoch(epoch, loader, iters, tokenizer, lm_config, start_step=0...

FILE: trainer/train_spo.py
  class AutoAdaptiveValueTracker (line 27) | class AutoAdaptiveValueTracker:
    method __init__ (line 29) | def __init__(self, rho_mode='kl', rho_const=0.9, D_half=0.06, clip_low...
    method get_baselines (line 40) | def get_baselines(self, batch_size):
    method compute_rho (line 44) | def compute_rho(self, cur_mean_logprob):
    method update (line 53) | def update(self, rewards, cur_logprobs=None, response_masks=None):
  function calculate_rewards (line 69) | def calculate_rewards(prompts, responses, reward_model, reward_tokenizer):
  function spo_train_epoch (line 131) | def spo_train_epoch(epoch, loader, iters, ref_model, reward_model, rewar...

FILE: trainer/train_tokenizer.py
  function get_texts (line 11) | def get_texts(data_path):
  function train_tokenizer (line 18) | def train_tokenizer(data_path, tokenizer_dir, vocab_size):
  function eval_tokenizer (line 87) | def eval_tokenizer(tokenizer_dir):

FILE: trainer/trainer_utils.py
  function get_model_params (line 18) | def get_model_params(model, config):
  function is_main_process (line 31) | def is_main_process():
  function Logger (line 35) | def Logger(content):
  function get_lr (line 40) | def get_lr(current_step, total_steps, lr):
  function init_distributed_mode (line 44) | def init_distributed_mode():
  function setup_seed (line 54) | def setup_seed(seed: int):
  function lm_checkpoint (line 63) | def lm_checkpoint(lm_config, weight='full_sft', model=None, optimizer=No...
  function init_model (line 119) | def init_model(lm_config, from_weight='pretrain', tokenizer_path='../mod...
  class SkipBatchSampler (line 134) | class SkipBatchSampler(Sampler):
    method __init__ (line 135) | def __init__(self, sampler, batch_size, skip_batches=0):
    method __iter__ (line 140) | def __iter__(self):
    method __len__ (line 155) | def __len__(self):